Re: C++ PATCH for c++/89686 - mixing init-capture and simple-capture in lambda

[Marek Polacek]

On Tue, Mar 12, 2019 at 04:07:56PM -0400, Jason Merrill wrote:
> On 3/12/19 3:59 PM, Marek Polacek wrote:
> > As Barry explained in the PR, lambda capture is one of
> > 
> >    simple-capture ...[opt]
> >    ...[opt] init-capture
> > 
> > where init-capture requires an initializer.  Here we have
> > 
> >    [...xs...]
> > 
> > which is ill-formed as it's mingling both of these.
> > 
> > Bootstrapped/regtested on x86_64-linux, ok for trunk?  Or should I defer to 
> > GCC
> > 10?
> > 
> > 2019-03-12  Marek Polacek  <pola...@redhat.com>
> > 
> >     PR c++/89686 - mixing init-capture and simple-capture in lambda.
> >     * parser.c (cp_parser_lambda_introducer): Give error when combining
> >     init-capture and simple-capture.
> > 
> >     * g++.dg/cpp2a/lambda-pack-init2.C: New test.
> > 
> > diff --git gcc/cp/parser.c gcc/cp/parser.c
> > index f95111169ed..d5d8f364752 100644
> > --- gcc/cp/parser.c
> > +++ gcc/cp/parser.c
> > @@ -10721,6 +10721,15 @@ cp_parser_lambda_introducer (cp_parser* parser, 
> > tree lambda_expr)
> >         {
> >           cp_lexer_consume_token (parser->lexer);
> >           capture_init_expr = make_pack_expansion (capture_init_expr);
> > +         if (init_pack_expansion)
> > +           {
> > +             /* We'd already seen '...' so we were expecting an
> > +                init-capture.  But we just saw another '...' which
> > +                would imply a simple-capture.  */
> > +             error_at (capture_token->location,
> > +                       "combining init-capture and simple-capture");
> 
> That diagnostic seems a bit obscure, how about something like "too many
> %<...%> in lambda capture"?

Yup, sounds better.

> Or perhaps check to see if there's an initializer after the ..., and
> complain about the second ... if so, or the first ... if not.

I tried but we already give diagnostic for [...xs=xs...] or [...xs=...xs]
and similar.  But let's at least use a proper location for the '...'.

Bootstrapped/regtested on x86_64-linux, ok for trunk?

2019-03-12  Marek Polacek  <pola...@redhat.com>

        PR c++/89686 - mixing init-capture and simple-capture in lambda.
        * parser.c (cp_parser_lambda_introducer): Give error when combining
        init-capture and simple-capture.

        * g++.dg/cpp2a/lambda-pack-init2.C: New test.

diff --git gcc/cp/parser.c gcc/cp/parser.c
index f95111169ed..69294e20191 100644
--- gcc/cp/parser.c
+++ gcc/cp/parser.c
@@ -10719,8 +10719,17 @@ cp_parser_lambda_introducer (cp_parser* parser, tree 
lambda_expr)
 
          if (cp_lexer_next_token_is (parser->lexer, CPP_ELLIPSIS))
            {
+             location_t eloc = cp_lexer_peek_token (parser->lexer)->location;
              cp_lexer_consume_token (parser->lexer);
              capture_init_expr = make_pack_expansion (capture_init_expr);
+             if (init_pack_expansion)
+               {
+                 /* We'd already seen '...' so we were expecting an
+                    init-capture.  But we just saw another '...' which
+                    would imply a simple-capture.  */
+                 error_at (eloc, "too many %<...%> in lambda capture");
+                 continue;
+               }
            }
        }
 
diff --git gcc/testsuite/g++.dg/cpp2a/lambda-pack-init2.C 
gcc/testsuite/g++.dg/cpp2a/lambda-pack-init2.C
new file mode 100644
index 00000000000..cd31f3296e1
--- /dev/null
+++ gcc/testsuite/g++.dg/cpp2a/lambda-pack-init2.C
@@ -0,0 +1,19 @@
+// PR c++/89686
+// { dg-do compile { target c++2a } }
+
+template <typename... Ts>
+void foo(Ts... xs)
+{
+  int i = 10;
+  [...xs...]{}(); // { dg-error "too many ..... in lambda capture" }
+  [xs...]{}();
+  [...xs=xs]{}();
+  [i, ...xs...]{}(); // { dg-error "too many ..... in lambda capture" }
+  [i, xs...]{}();
+  [i, ...xs=xs]{}();
+}
+
+int main()
+{
+ foo(0, 1, 2);
+}